Importance of Precision

[proto57]

I recently watched the online Voynich documentary by the YouTube channel, "The Histocrat", and wrote about it in another forum post. In short, I found that it is well-produced for an amateur doc, and that like all secondary works, is based entirely on available information, and little or no original research. That is also understandable. And as usual, a great deal of it is from the very complete Voynich.nu site, with the Yale book, D'Imperio, and various other well-known sources being sprinkled in.

But as I also said, I think this work is a great example of the danger of the established sources, such as .nu, strongly projecting the opinions of the creator as known and factual. Anyone after that can't be blamed for not being able to tell when something is opinion, and then they further expound on this opinion, further cementing them as fact, and closing off any chance people will explore any possibles outside of "1420 Genuine European Herbal". So if it is not that, it will never be discovered... or be harder to. But then, that is the point of these sources doing this, they don't want that paradigm questioned. But why and how is another story.

What I wanted to do here was show one clear example of this danger, from this video. The section is on the radiocarbon testing, and starts about 1:02:

https://youtu.be/UZEHBkBlalc?t=3722

After describing the testing (and not identifying Mr. Hodgins in the pictures!), it is stated,

"All four samples [of the parchment] gave matching results. With 95% certainty, we now know that the parchment of the Voynich Manuscript originated somewhere between 1404 and 1438".

Of course this is incorrect, because all four sample actually gave very DIFFERENT results. Those results, over 60 years apart, were then "combined" on the "assumption" that the Voynich was assembled within a few short years. As I wrote on my "Myths" blog post:

The published range is actually a conclusion determined by combining the very different results of the four samples tested. But when looked at separately, as would have been done if not found bound together, nor assumed to be made as the same time, the results show they could be 50 to 60 years apart. And taking into account the extremes of the error range of the samples, they actually could date to as much as 132 years apart:

Folio 8: 490±37, which works out to 1423 to 1497
Folio 26: 514±35, which works out to 1401 to 1471
Folio 47: 506±35, which works out to 1409 to 1479
Folio 68 (cleaned): 550±35, which works out to 1365 to 1435

The only reason we happen to have those results at all was because I quickly snapped the slide presented by Mr. Hodgins in 2012, at his lecture in Frascati (pats self on back). But the point I always make is that we should start with those factual results, and work from there. This "combining" would not have been done this if the samples were, for instance, an Italian land deed, a German letter, a British law, and a Spanish court record. We would not then say that all four of those items "date somewhere between 1404 and 1438, with 95% certainty", because they would be treated separately.

OK so here is what happened in the documentary, the narrator later states,

"It also makes it less likely that the manuscript was a fraud, perpetrated by Voynich himself, as it would have been extremely difficult to gather such a quantity of matching medieval parchment, from which to fashion such a fraud."

When we look at that statement, there is first, an affirmation based on an error ("matching"). But also, looked at a different way, when corrected for that error, there is an implied admission that if the 60 year spread of actual parchment dates was known to the narrator, that it would not only be a weakness to genuine, but even, a support for forgery.

An allegory to this effect might be if it were said, "Fred can't be guilty because they never found a weapon on him", when the narrator is unaware that a weapon was actually found on Fred. The "pass" the narrator then becomes, by his own admission, a support of guilt.

In any case, this is only one example of literally dozens in this video, and for that matter other ones, and also books, articles, blogs and forums, which show that when one relies on poor information and/or the opinions of others, the conclusions based on these can only be poor ones. But also, within the secondary conclusions, when one knows the source used, and reads... in the author or narrators own words... the reasoning used on that source, it becomes clear that if they had a full understanding of what is actually known, and not known, they probably would realize there is a problem in paradise, that the Voynich cannot be what they have been told to believe it is. I mean, I can see the wheels of common sense, critical thinking, in these accounts... and can't blame them for not realizing the information they base it on, is incorrect.

And that gives me hope, of a sort. It's not the reasoning process that is at fault, it is the tools they are given to work with, that is the problem.

[R. Sale]

Precision is the key, relatively speaking. C-14 tests are not that accurate, as seen by the data. Yet I believe it has been said that we are lucky to be where we are in the chronology because precision is worse elsewhere. And there is a certain commonality to the data results.

The widely promoted dating range for the four VMs parchment samples (1404-1438) is based on the assumption that all four samples were from a single source, and the variation was in the data - not in the sample. The distance between the different sample dates is not so great that it can rule out this possibility.

The single source assumption is just one way to interpret the data. What if there was a second source of parchment? The two source assumption may be the second best assumption. It may be less probable, but it is just as viable as a real possibility. The single source assumption is not the only possible assumption.

Or perhaps we should exclude the sample that was "cleaned". If this is a two source system the chronological presumptions will need to be changed. Possibly pushing the earliest possible 'authentic MS' assumption forward to 1450 or beyond. It makes some difference historically.

Or else it's just a lucky accident for some flamboyant, fraudulent, fatally-flawed forger. No precision there - or is there?

[proto57]

Precision is the key, relatively speaking. C-14 tests are not that accurate, as seen by the data. Yet I believe it has been said that we are lucky to be where we are in the chronology because precision is worse elsewhere. And there is a certain commonality to the data results.

What do you mean by "commonality"?

The widely promoted dating range for the four VMs parchment samples (1404-1438) is based on the assumption that all four samples were from a single source, and the variation was in the data - not in the sample. The distance between the different sample dates is not so great that it can rule out this possibility.

There can be many ways to interpret data of course, but in my opinion it becomes useless to determining the nature of the tested item when too many variables are allowed, and those variables are moved to "make" the item what one "wants" it to be. So yes, one can say the "variation was in the data", and not the sample, but that means the data was just interpreted as inaccurate by just enough to "make" the book made from one source of parchment. That is the same as saying that the data is wrong, and should be ignored. But then what is the point of ever testing a multiple of samples from any assembled object, if one can so ignore the results anyway?

I see this done all the time, with all sorts of tests. And the testers and the interpreters of that data often seem unaware that they are doing it, but it can be found in almost all accounts of any forensic investigation. The Yale book is rife with such things, including what I consider the classic line about the "very unusual for the time" foldouts:

"The quantity and size of the foldouts in the Voynich Manuscript are very unusual for the time period; it is rare to find so many large pieces of parchment folded into a single textblock, and this seems to indicate authenticity: In the twentieth century it would be quite difficult to find this many large sheets of genuine medieval parchment in order to produce a forgery."- Zyats et al, Yale Voynich publication.

I see that as a scientist realizing that the data is indicating forgery; but that they know what the desired result is, "genuine and old". So they try and counter the observation with a statement that happens to be false, but sort of kind of maybe will sound sort of OK to any reader... and more importantly, the publisher of the book the manuscript is about, who also happens to be the owner of the manuscript. Would that incorrect disclaimer about it "seems to indicate authenticity" be in an independent report? I doubt it.

This is a parallel to any claim the C14 data proves authenticity, to me. No, it implies construction from multiple-aged parts, which undermines authenticity, because that is not normally found. There is a bit of a difference from the foldout claim, though... rather than say multiple aged parchment is "an indication of authenticity", they simply change the dating, by "combining" those varied dates. One cannot make the foldouts disappear in this way, however... they can't be ignored, or cut out and discarded, like those dates can be, so they are painted into a corner, and must try to counter them.

I see this constantly: Data is very often not seen as the valuable, identifying tool that is is supposed to be, but more as an impediment to certifying items as what their patrons want them to be. What often follows, and followed here, is that these "impediments" are therefore removed (the Purple Cow Syndrome), or corrupted (adjusted to fit, or otherwise) between the gathering of the data and the report. And if neither of those can be done, "qualifying phraseology" is applied in an attempt to counter that undesired data, such as the above, bizarre, foldout line from Yale.

It was fun, and gratifying for me, to read the recent book about the Hoffman forgeries. It is a collection of reports, letters and essays about his forged "Oath of a Freeman", and several of them are from before they realized it WAS a forgery. The effects I describe above are all over these "professional", supposedly scientific papers. You can see the machinery of the investigator first realizing the problems with the "Oath", but then dismissing them, rather than "letting" them tell them what these problems really were: That it was a fake.

The single source assumption is just one way to interpret the data. What if there was a second source of parchment? The two source assumption may be the second best assumption. It may be less probable, but it is just as viable as a real possibility. The single source assumption is not the only possible assumption.

True, we could do that. But I would say that it would be, again, looking for ways to explain the data to fit any "desired" results, rather than the opposite, which I am arguing here SHOULD be done: Let the data tell us how the thing was constructed, and with what age materials, rather than us manipulate the data to fit a result we choose.

Or perhaps we should exclude the sample that was "cleaned". If this is a two source system the chronological presumptions will need to be changed. Possibly pushing the earliest possible 'authentic MS' assumption forward to 1450 or beyond. It makes some difference historically.

Again... I'd make the same points for this. Why should we be "pushing" the date anywhere, and now, by removing a result? Rather, I think we should let it tells us what it tells us: That the Voynich is made from a parchment assortment which varies in age about 60 years or more.

Or else it's just a lucky accident for some flamboyant, fraudulent, fatally-flawed forger. No precision there - or is there?

Well I would call the choice of parchment with a 60 year age range "unlucky for the forger", not lucky. He/she didn't predict C14, and that invention was also "unlucky" for them. But if they lived, or watching "from Above", they need not worry... for what was most "lucky" for the forger has been the huge effort and energy expended, by all the work's defenders, for over 108 years, to protect it from all it's anachronisms, anomalies and inconsistencies, no matter how bad they actually are: If something looks anachronistic, say it is coincidence or paradiolia. If the data does not fit, alter the data to fit. If neither can be done, remove them. If you can't remove the "offending" data or item, ignore it. And if none of the above can be done, just make stuff up to explain it away, as was done with the foldouts, and really, so many other things.

Nobody will check, and if they do, and find out it is wrong, simply ignore that, too. That is, they do anything to be able to support this house of cards, and science be damned.

[R. Sale]

I do believe we agree on much of this. It's a game of self-promotion. And if data and facts get in the way, they can be manipulated or ignored. That's what WMV was doing, calling his book a 'Roger Bacon' manuscript. Either he knew nothing about it, or he knew it was totally false.

By commonality I meant that all results were pretty much in the 1400s, They are in the same century. So even if the four samples came from four separate sources, there are some unknown constraints that limit the results to a single century - without knowing about C-14 dating yet to come.

My concern with the importance of precision has to do with intellectual honesty. If data is only used for self-promotion, then precision does not matter. The data is what it is. It is the assumption that determines how the data is used. In a situation where multiple interpretations exist, selecting the most favorable possibility and disallowing the other interpretations is problematic.

That is the problem with single source data average. If there was a second source, the average is not valid. The data does not reveal how many sources there were. It only tries to date them. And this is only for the parchment. It says nothing definitive about the creation of the manuscript.

Precision can also be applied to terminology. Precision can be measured in the VMs by its connections with tradition and its capacity to provide useful interpretation. The use of a nebuly line in the VMs cosmos and VMs critter (f80v), much like the example in the "Berry Apocalypse" shows that tradition is either merely copied or potentially understood. The parts of the VMs cosmos, in comparison to the 'Oresme' example in BNF Fr. 565, indicates that the VMs artist knew the tradition, but chose not to copy it, when copying would seem to be the 'forger's' better choice.

In the matter of the the obscure, heraldic fur called papelonny, the artist indicates familiarity through choice and manipulation. A copy is a copy. Manipulation indicates understanding. Does WMV ever explain any part of the VMs based on his accurate understanding of tradition? IMO, there seems to be a disconnect. A higher level of knowledge is needed to manipulate tradition than to copy it, and it's not a copy, while the premise behind the WMV explanation is without justification.

[proto57]

Thank you for the discussion, Richard, as usual.

Rich.

[Dion2]

Folio 8: 490±37, which works out to 1423 to 1497
Folio 26: 514±35, which works out to 1401 to 1471
Folio 47: 506±35, which works out to 1409 to 1479
Folio 68 (cleaned): 550±35, which works out to 1365 to 1435

This is chastening for me. I've always accepted the range as 1404-1438, but obviously should have been considering 1365-1479.

And I would agree with all you say about .nu Any old hand comes eventually to realise just how much bias and skewed information is being presented there in a tone suggesting it is established truth. The very fact of an amateur's presenting his preferred ideas as if those of some authority should have woken me up a lot earlier, but it has taken a while.